Anda di halaman 1dari 8

194

Semantic memory and the brain: structure and processes Alex Martin* and Linda L Chao
Recent functional brain imaging studies suggest that object concepts may be represented, in part, by distributed networks of discrete cortical regions that parallel the organization of sensory and motor systems. In addition, different regions of the left lateral prefrontal cortex, and perhaps anterior temporal cortex, may have distinct roles in retrieving, maintaining and selecting semantic information.
Addresses Laboratory of Brain and Cognition, National Institutes of Mental Health, Bethesda, MD 20892, USA *e-mail: alex@codon.nih.gov Correspondence: Alex Martin Current Opinion in Neurobiology 2001, 11:194201 0959-4388/01/$ see front matter 2001 Elsevier Science Ltd. All rights reserved. Abbreviations fMRI functional MRI LIPC left inferior prefrontal cortex MRI magnetic resonance imaging MT middle temporal

these broadly defined areas. In this review, we will highlight studies that have shed new light both on the neural structures involved in representing object-specific features and attributes, and on the processes involved in retrieving and manipulating this information.

The structure of semantic representations


An old idea in behavioral neurology [6,7] is that object concepts are defined by sensory and motor attributes and features acquired during experience. Within this framework, many have suggested that object concepts may be represented in the brain as distributed networks of sensory, motor and/or more abstract functional information [814]. Investigations of word-generation have provided one body of evidence suggesting that information about different object features may be stored in different regions of the cortex. These studies have shown that the ventral and lateral regions of posterior temporal cortex can be differentially engaged depending on the type of information retrieved [15]. For example, we reported that asking subjects to generate the name of an action typically associated with an object activated the posterior region of the left middle temporal gyrus, just anterior to sites active during motion perception, whereas generating a color word activated the ventral temporal lobes (bilateral fusiform gyrus) just anterior to sites active during color perception [15]. Activation of the fusiform gyrus when subjects retrieve color word associates has recently been replicated in two additional studies [16,17]. Moreover, activation in a similar region has been reported during the spontaneous generation of color imagery in auditory color-word synaesthetes [18] and in normal individuals during color imagery [19]. These data should be viewed in the context of over 20 reports in the literature that have found greater activity in the left posterior middle temporal gyrus during action word generation to visually and auditorily presented words and pictures (including re-analyses [20,21] of the original Petersen et al. study [22]; see [23] for a review). More recently, this corpus of findings has been extended to subjects tested across a wide variety of native languages [24,25], as well as to bilingual individuals responding in their native and second languages [26]. Taken together, these word generation data are consistent with the idea that information about object-specific features may be stored within the same neural systems that are active during perception.

Introduction
The domain of semantic memory consists of stored information about the features and attributes that define concepts and the processes that allow us to efficiently retrieve, act upon and produce this information in the service of thought and language. Before the advent of functional brain imaging, our knowledge of the neural bases of semantic memory was dependent on studies of patients with brain injury or disease. These investigations identified at least two brain regions, particularly in the left hemisphere, that play a critical role in semantic memory. Patients with damage to the left prefrontal cortex (LPC) often have difficulty retrieving words in response to specific cues (e.g. words beginning with a specific letter, the names of objects belonging to a specific semantic category), even in the absence of a frank aphasia [1]. This suggests that the LPC plays a general, albeit crucial, role in retrieving lexical and semantic information. Patients with damage to the temporal lobes often have difficulty naming objects and retrieving information about object-specific characteristics [24]). This suggests that object-specific information may be stored, at least in part, in the temporal lobes. Consistent with the clinical literature, early functional brain imaging studies of semantic processing revealed activity in broad expanses of the left prefrontal, parietal and posterior temporal lobes, commonly including ventral and lateral regions of temporal cortex [5]. Results from recent studies have begun to reveal specific functions and processes subserved by smaller regions within each of

Distributed representations of object categories


Another body of evidence that object concepts may be represented by distributed feature networks comes from studies contrasting patterns of neural activity associated with performing the same type of task (e.g. naming) with

Semantic memory and the brain: structure and processes Martin and Chao

195

Figure 1 Group fMRI data (N = 7), superimposed on a group-averaged structural MRI, showing category-related activations associated with silently naming pictures of animals and tools. Regions showing greater activity for naming animals than for naming tools (1: lateral fusiform gyrus; 4: right superior temporal sulcus) are shown in white. Regions showing greater activity for naming tools than for animals (2: medial fusiform gyrus; 3: left middle temporal gyrus/inferior temporal sulcus; 5: left ventral premotor cortex) are shown in black. y = 41

z = +36 5

z = 8

4 1 1 2 2 3 1 L 1 2 2 3

L Animals > tools Tools > animals


Current Opinion in Neurobiology

different categories of objects. A common feature of all concrete objects is their physical form. Evidence is accumulating that suggests that all object categories elicit distinct patterns of neural activity in regions that mediate perception of object form (the ventral occipitotemporal cortex). Moreover, the locations of these category-related activations, especially for objects defined primarily by their visual form-related features such as animals, faces and landmarks, appear to be consistent across individual subjects and processing tasks. Information about how objects move through space and patterns of motor movements associated with their use are other features that could aid object recognition and identification. This would be especially true for categories of man-made objects such as tools that have a more variable mapping between their name and their visual form than a category such as four-legged animals. Thus, access to these additional features may be required to identify them as unique entities. Here again, evidence is accumulating that naming and identifying objects with motion-related attributes activate areas close to regions that mediate perception of object motion (the posterior region of the lateral temporal lobe), with different patterns of activity associated with biological and man-made objects. Similarly, naming manipulable man-made objects selectively activates areas close to regions active during object manipulation.

Finally, these regions are active not only when objects are viewed and named: answering written questions and imagining them also elicits activity in these regions. Such findings are consistent with the idea that these activations reflect retrieval of stored information about object-specific attributes and features.
Ventral occipitotemporal cortex and the representation of object form

A number of investigators have found that distinct regions of ventral temporal cortex show differential responses to different object categories. In a series of studies [27], we found greater activity in the lateral region of the fusiform gyrus for animal than for tool stimuli. This finding was consistent among different tasks, including viewing, naming, and matching pictures, and answering written questions about object features. In contrast, the medial fusiform gyrus was more active for tools than for animals (see Figure 1). A similar, but not identical pattern of activation was found for viewing faces (in the lateral fusiform) relative to viewing houses (the medial fusiform) [27]. Other investigators have also reported face-related activity in the lateral region of the fusiform gyrus [2830] and house-related activity in more medial regions, including the fusiform and lingual gyri [31] and parahippocampal cortex (especially for landmarks [32]).

196

Cognitive neuroscience

Fine-grained distinctions were documented between each of these categories (i.e. animals, tools, houses, and faces ([27,33], although see [34]). Nevertheless, the data suggest that activations associated with animate objects or living things (i.e. animals, faces) cluster in the more lateral aspect of the fusiform gyrus, whereas activations associated with inanimate or man-made objects (i.e. tools, houses) cluster in the more medial aspect of the fusiform gyrus. Thus, category-related activations may conform to an animate/inanimate organizational scheme as most recently suggested by Caramazza and Shelton [35]. However, findings by Ishai et al. [36], using an arbitrary category that has no evolutionary significance, casts doubt on this possibility. In this study, subjects viewed and performed matching tasks with pictures of chairs, faces and houses. Rather than falling medial to the activation associated with faces (i.e. in the medial fusiform region, more active for houses than faces, and more active for tools than animals), the main peak of activity for chairs was lateral to the face-responsive area, falling in the inferior temporal gyrus. These findings suggest that different object categories elicit activity in different regions of the ventral temporal cortex, as defined by the location of their peak activation. In addition, the typological arrangement of these peaks was consistent across subjects and tasks. However, it is important to note that the activity associated with each object category was not limited to one region, but rather involved a relatively large expanse of ventral cortex. Moreover, rather than a single peak, this activity was best characterized by a complex pattern of peaks and valleys distributed over much of ventral temporal and, as discussed below, occipital cortex as well. This suggests that the representations of different object categories are distributed and overlapping. Activation of medial and ventrolateral occipital cortex for animals compared with tools has been reported using object-naming [27,37], picture-matching [38,39], and word-reading [39] tasks. Converging evidence has come from a report [40] that patients with unilateral lesions of the medial and ventral occipital cortex are more impaired at recognizing animals than tools and famous faces. Because the activations occurred so early in the visual-processing stream, we suggested that they might reflect top-down activation from more anterior sites. This may occur whenever detailed information about visual features is needed to identify a specific object [27,37,41]. However, evidence about the onset times of occipital and temporal activity will be needed to determine whether these occipital activations represent bottom-up or topdown effects (see [42,43] for additional evidence for category-related activations in occipital cortex). More direct evidence for the distributed nature of object representations in the ventral temporal cortex comes from single-cell recordings from intracranial depth electrodes implanted in patients with epilepsy [44]. Here, Kreiman

et al. recorded from the medial temporal cortex (i.e. the entorhinal cortex, hippocampus and amygdala), which receives major inputs from the ventral temporal regions identified in the neuroimaging studies described above. Neurons were identified that showed highly selective responses to different object categories including animals, faces, and houses. Moreover, the responses of the neurons were category-specific rather than stimulus-specific. That is, animal-responsive cells responded to all pictures of animals, rather than to one picture or a select few. Studies reporting similar patterns of activation when subjects view and imagine objects from different object categories provide further support that the responses in these regions are driven by stored object information. For example, OCraven and Kanwisher [45] reported that the same regions were active when subjects both viewed and imagined famous faces and known landmarks. Similarly, Ishai et al. [46] found a correspondence between the brain regions active during viewing and imagining faces, houses and chairs. Finally, Kreiman et al. [44] reported that the majority of category-selective neurons (88%) in their study also responded selectively when the patients were asked to imagine these objects. Taken together, these data suggest that ventral occipitotemporal cortex may be best viewed not as a mosaic of discrete category-specific areas, but rather as a lumpy feature-space, representing stored information about features of object form shared by members of a category [41,47,48]. A feature-based model can accommodate the observation that an arbitrary category such as chairs elicited a pattern of neural activity distinct from other object categories (i.e. faces and houses). Clearly, it would be difficult, as well as unwise, to argue that there is a chair area in the brain. There are simply too many categories, and too little neural space to accommodate discrete, category-specific modules for every category. In fact, there is no limit on the number of object categories. Feature-based models can provide the flexibility needed to represent an infinite variety of objects. How this feature space is organized, and why its topological arrangement is so consistent from one subject to another, are critical questions for future investigations.
Lateral temporal cortex and the representation of object motion

A number of laboratories, using a variety of paradigms with pictures and words, have reported that tools elicit greater activity in the left posterior middle temporal gyrus than animals and other object categories [27,37,39,49]. Moreover, the active region is just anterior to area MT (middle temporal) and slightly posterior to, or overlapping with, the region active in the verb-generation studies discussed above (see [23] for a review). Damage to this region has been reported to selectively impair tool recognition and naming [40]. In contrast, naming animals [27] and viewing faces [27,28,30] elicits greater activity in the superior temporal sulcus in approximately half the subjects

Semantic memory and the brain: structure and processes Martin and Chao

197

Figure 2 Schematic representation of (a) ventral and (b) lateral surfaces of the brain. Shown are the approximate locations of regions where information about object form, motion and object-use-associated motor patterns may be stored. Information from an increasing number of sources may be integrated in the temporal lobes, with specificity increasing along the posterior to anterior axis. Specific regions of the LIPC and the polar region of the temporal lobes may be involved differentially in retrieving, monitoring, selecting and maintaining semantic information.

(a)

(b) Motor Selection

Production

Biological motion artifact

Form Semantic working memory system Generic Unique


ce/ ergen Conv tion ra integ
Current Opinion in Neurobiology

Visual detail

tested (see Figure 1). This region is of particular interest because of its association with the perception of biological motion in monkeys [50] as well as humans [51,52]. As suggested for the ventral temporal cortex, neurons in the lateral temporal cortex may also be tuned to features that objects within a category share. The nature of these features is unknown; however, on the basis of its anatomical proximity to visual motion processing areas, this region may be tuned to features of motion associated with different objects. In support of this conjecture, increased activity in posterior lateral temporal cortex has been found when subjects view static pictures of objects that imply motion [53,54] and when subjects focused attention on the direction of eye gaze [55]. Investigation of the differences in the properties of motion associated with biological and man-made objects may provide clues to the organization of this region.
Ventral premotor cortex and the representation of use-associated motor movements

when objects are grasped, but also when the animals view objects that they have previously manipulated [61]. The ventral premotor activation noted in the human neuroimaging studies may reflect a similar process. Alternatively, this activation may reflect action-planning [62]. Studies of patients with left premotor lesions will be needed to determine whether this region is necessary for naming and retrieving information about tools.

Representation of the subordinate level and unique objects


A common characteristic of the studies described above was that object categories were represented by items named at the basic level (i.e. house, face, chair, dog, hammer) rather than at the subordinate or unique-object level (e.g. the White House, Marilyn Monroe, rocking chair, collie, sledgehammer). The most commonly studied category of unique entities has been famous faces. These investigations typically have observed activity in the anterior middle temporal gyrus [63,64,65] and temporal pole [63,66]. Evidence that activity in anterior temporal regions may extend to other classes of objects was reported by Gauthier et al. [67], who used pictureword matching tasks on the basic and subordinate levels (although see [68]). These findings are consistent with Damasios assertion that the anterior regions of the temporal lobes are critical for retrieving information about unique entities [66,69]. Why this should be so is not clear. It can be assumed, however, that naming unique entities and making subordinate-level distinctions require access to more information than basic-level identification. Thus, one possibility is that the temporal lobe object representation system is organized hierarchically, with increasing convergence and integration of information occurring along its posterior to anterior axis (see Figure 2).

If activations associated with different object categories reflect stored information about object properties, one would expect tools to elicit activity in motor-related regions. Several laboratories have reported this association. Specifically, greater activation of left ventral premotor cortex has been found for naming tools relative to naming animals [37,56,57] (see Figure 1), viewing pictures of tools compared with viewing pictures of animals, faces and houses [57], and generating action words to tools [58]. As with studies of object form and object color, mental imagery (e.g. imaging manipulating objects with the right hand) also results in ventral premotor activation [59,60]. Electrophysiological studies have identified cells in monkey ventral premotor cortex that respond not only

198

Cognitive neuroscience

Working with semantic representations


As noted previously, performing semantic tasks commonly activates a wide expanse of left lateral prefrontal cortex. More recently, evidence has accumulated suggesting that an anterior and inferior prefrontal region (roughly equivalent to Brodmanns Area BA 47 and the inferior aspect of BA 45) may be involved selectively in semantic processing. Specifically, as suggested by Gabrieli and co-workers, and Wagner [70,71,72], this region may serve as a semantic working memory system responsible for retrieving, maintaining, monitoring and manipulating semantic representations stored elsewhere. Evidence supporting this view comes from studies showing that the left inferior prefrontal cortex (LIPC) is more active when subjects make semantic judgements to words than when they make nonsemantic judgements to the same words [73], even when the tasks are equated for difficulty [72,74]. Additional evidence comes from studies that find repetition-related decreases (and more efficient processing) in LIPC activity with repeated performance on semantic, but not nonsemantic, tasks [74,75]. Whereas the studies mentioned above define semantic processing by abstract/concrete classification of single words, a recent study [76] has extended the selective activation of the LIPC to sentence-level semantic processing as well. Specifically, judging whether pairs of sentences have the same meaning activates LIPC when meaning is determined by synonyms (e.g. the car is in the garage; the auto is in the garage) relative to when meaning is determined by syntax (e.g. the pool is behind the gate; behind the gate is the pool). Although these findings are broadly consistent with the idea that this region may subserve a number of workingmemory-like processes, the precise role of LIPC is difficult to determine from these reports. In contrast, ThompsonSchill and colleagues have focused on the role of LIPC in mediating a single processing component: selection among competing alternatives. Support for this idea has been obtained by demonstrating that LIPC activity is modulated by selection demands. For example, generating verbs to nouns that typically have multiple competing responses (high selection demand; e.g. turn or spin or roll to wheel) produces more LIPC activity than performing the same task with nouns that have more stereotypic verb responses (low selection demand; e.g. cut to scissors) [77]. Moreover, patients with LIPC lesions are impaired when required to generate verbs to nouns with high, but not low, selection demand [78]. Because high-selection-demand tasks also place greater demands on retrieval, these findings may reflect retrieval demands, rather than selection demands per se. To distinguish between these alternatives, subjects were asked to generate different features (e.g. action word, color word) to the names of concrete objects [79]. The authors reasoned that in both conditions, a common set of semantic features would be retrieved, but different attributes would be selected depending on the task demand. As in previous

reports using semantic classification [74,75] and word generation [22] tasks, repeated performance of the same task (i.e. repeated generation of an action word to the same concrete noun) resulted in decreased activity in the LIPC. However, crossing the tasks (i.e. generating an action word to a concrete noun when the subjects had previously generated a color word to that noun) produced increased LIPC activity, even though the same concept and set of attributes were retrieved under both conditions. Thus, when recently selected information was irrelevant, selection demands and LIPC activity increased, whereas when recently selected information was relevant, selection demands and LIPC activity decreased. In contrast, temporal lobe activity decreases when the same concept is retrieved, regardless of the type of feature that has to be selected. Taken together, these findings are consistent with the idea that LIPC is involved in the selection among competing semantic features (color and action object associates) stored in temporal cortex. Interestingly, the location of the selection-related LIPC activity is actually in the dorsal aspect of BA 44, placing it posterior and superior to the semantic working memory region identified by Gabrieli, Wagner and colleagues [70,71,72]. This suggests that the process of selecting among competing alternatives may occur in a different cortical region than other working memory processes. Although helping to clarify the role of different subregions of the LIPC, this story is complicated by the fact that phonological processing tasks also activate dorsal BA 44 (as well as the more inferior portion of BA 44; see [72] for a review). One possibility is that the phonological tasks also require selection. Alternatively, BA 44 may be fractionated functionally into two regions, with the inferior portion involved in phonological processing especially transforming orthographic to phonologic representations for speech production [80] and the superior portion engaged in selection processes (see Figure 2). Finally, recent evidence suggests that the polar region of the left temporal lobe may also contribute to the overall functioning of a semantic working memory system. Mummery et al. [81] tested normal subjects and patients with semantic dementia on a semantic association task. Although the patients had extensive anterior temporal lobe atrophy, they showed either normal or heightened activity in subregions of the left anterior temporal lobe (and the LIPC). In contrast, whereas normal individuals showed strong activation of the ventral and lateral regions of the posterior left temporal cortex, these patients failed to activate this region. Thus, within the context of the material reviewed here, normal functioning of the polar region of the left temporal lobe may provide top-down modulatory input necessary for successful retrieval of semantic representations stored in posterior regions (see Figure 2). In line with this idea, lesions of the anterior temporal cortex in

Semantic memory and the brain: structure and processes Martin and Chao

199

monkeys have recently been shown to interfere with maintaining items in visual working memory [82]. Clearly, the role of different regions of left prefrontal cortex in semantic memory processes and, more generally, in lexical retrieval and production, remain to be specified. In particular, the idea that the polar region of the left temporal lobe is part of a semantic working memory network awaits further study.

9.

Warrington EK, McCarthy R: Categories of knowledge: further fractionation and an attempted integration. Brain 1987, 110:1273-1296.

10. Warrington EK, Shallice T: Category-specific semantic impairments. Brain 1984, 107:829-854. 11. Farah MJ, McClelland JL: A computational model of semantic memory impairment: modality specificity and emergent category specificity. J Exp Psychol 1991, 120:339-357. 12. Damasio AR: Category-related recognition deficits as a clue to the neural substrates of knowledge. Trends Neurosci 1990, 13:95-98. 13. Caramazza A, Hillis AE, Rapp BC, Romani C: The multiple semantics hypothesis multiple confusions. Cogn Neuropsychol 1990, 7:161-189. 14. Humphreys GW, Riddoch MJ: On telling your fruit from vegetables: a consideration of category-specific deficits after brain damage. Trends Neurosci 1987, 10:145-148. 15. Martin A, Haxby JV, Lalonde FM, Wiggs CL, Ungerleider LG: Discrete cortical regions associated with knowledge of color and knowledge of action. Science 1995, 379:649-652. 16. Chao LL, Martin A: Cortical representation of perception, naming, and knowing about color. J Cogn Neurosci 1999, 11:25-35. 17. Wiggs CL, Weisberg J, Martin A: Neural correlates of semantic and episodic memory retrieval. Neuropsychologia 1999, 37:103-118.

Conclusions
Distributed networks of discrete cortical regions are active during object processing. The distribution of these regions varies as a function of semantic category. The same regions are active, at least in part, when objects from a category are recognized, named, imagined, and when reading and answering questions about them. Critical questions for future research will be to clarify the precise role of these regions in object semantics and how are they influenced by experience. In addition, it has yet to be determined how the lexicon is organized and how lexical representations are linked to the semantic feature networks described here. Finally, little is known about the neural representations of nonfeatural semantic object information and abstract concepts. Nevertheless, the findings reviewed here suggest that we are beginning to make progress in identifying the distributed cortical networks associated with semantic object representations, and the networks underlying our ability to retrieve, select and operate upon them.

18. Paulesu E, Harrison J, Baron-Cohen S, Watson JDG, Goldstein L, Heather J, Frackowiak RSJ, Firth CD: The physiology of coloured hearing: a PET activation study of colour-word synaesthesia. Brain 1995, 118:661-676. 19. Howard RJ, Ffytche DH, Barnes J, McKeefry D, Ha Y, Woodruff PW, Bullmore ET, Simons A, Williams SCR: The functional anatomy of imagined and perceived colour. Neuroreport 1998, 9:1019-1023. 20. Raichle ME, Fiez JA, Videen TO, MacLeod AMK, Pardo JV, Petersen SE: Practice-related changes in human brain functional anatomy during non-motor learning. Cereb Cortex 1994, 4:8-26. 21. Fiez JA, Raichle ME, Balota DA, Tallal P, Petersen SE: PET activation of posterior temporal regions during auditory word presentation and verb generation. Cereb Cortex 1996, 6:1-10. 22. Petersen SE, Fox PT, Posner MI, Mintun M, Raichle ME: Positron emission tomographic studies of the cortical anatomy of singleword processing. Nature 1988, 331:585-589. 23. Martin A: Functional neuroimaging of semantic memory. In Functional Neuroimaging of Semantic Memory. Edited by Cabaza R, Kingstone A. Cambridge, Massachusetts: MIT Press; 2001:153-186. 24. Poline JB, Vandenberghe R, Holmes AP, Friston KJ, Frackowiak RSJ: Reproducibility of PET activation studies: lessons from a multicenter European experiment EU concerted action on functional imaging. Neuroimage 1996, 4:34-54. 25. Tatsumi IF, Fushimi T, Sadato N, Kawashima R, Yokoyama E, Kanno I, Senda M: Verb generation in Japanese a multicenter PET activation study. Neuroimage 1999, 9:154-164. 26. Klein D, Oliver A, Milner B, Zatorre RJ, Zhao V, Nikelski J: Cerebral organization in bilinguals: a PET study of ChineseEnglish verb generation. Neuroreport 1999, 10:2841-2846. Increased activity in the left posterior middle temporal gyrus, as well as in the left inferior and dorsolateral frontal cortex, parietal cortex and right cerebellum, was reported when bilingual subjects performed verb-generation tasks in their native (Mandarin Chinese) and second (English) language, suggesting shared neural substrates. 27. Chao LL, Haxby JV, Martin A: Attribute-based neural substrates in temporal cortex for perceiving and knowing about objects. Nat Neurosci 1999, 2:913-919. Consistent category-related activity in the ventral and lateral regions of the posterior temporal cortex was found across subjects and across multiple tasks using pictures of animals and tools. Similar patterns of category-related activity were also observed when subjects read and answered questions about animals and tools, suggesting that category-related activity in the posterior temporal cortex reflects stored information about objects. 28. Kanwisher N, McDermott J, Chun MM: The fusiform face area: a module in human extrastriate cortex specialized for the perception of faces. J Neurosci 1997, 17:4302-4311.

Acknowledgements
This work was supported by the National Institute of Mental Health Intramural Research Program.

References and recommended reading


Papers of particular interest, published within the annual period of review, have been highlighted as:

of special interest of outstanding interest


1. 2. 3. Baldo JV, Shimamura AP: Letter and category fluency in patients with frontal lobe lesions. Neuropsychology 1998, 12:259-267. Warrington EK: The selective impairment of semantic memory. Quart J Exp Psychol 1975, 27:635-657. Hart J, Gordon B: Delineation of single-word semantic comprehension deficits in aphasia, with anatomical correlation. Ann Neurol 1990, 27:226-231. Hodges JR, Patterson K, Oxbury S, Funnell E: Semantic dementia. Progressive fluent aphasia with temporal lobe atrophy. Brain 1992, 115:1783-1806. Demonet JF, Chollet F, Ramsay S, Cardebat D, Nespoulous JL, Wise R, Rascol A, Frackowiak R: The anatomy of phonological and semantic processing in normal subjects. Brain 1992, 115:1753-1768. Broadbent WH: A case of peculiar affection of speech with commentary. Brain 1878, 1:484-503. Lissauer H: A case of visual agnosia with a contribution to theory. Cogn Neuropsychol 1890, 5:157-192. Warrington EK, McCarthy R: Category specific access dysphasia. Brain 1983, 106:859-878.

4.

5.

6. 7. 8.

200

Cognitive neuroscience

29. McCarthy G, Puce A, Gore JC, Allison T: Face-specific processing in the human fusiform gyrus. J Cogn Neurosci 1997, 9:605-610. 30. Haxby JV, Ungerleider LG, Clark VP, Schouten JL, Hoffman EA, Martin A: The effect of face inversion on activity in human neural systems for face and object perception. Neuron 1999, 22:189-199. 31. Aguirre GK, Zarahn E, DEsposito M: An area within human ventral cortex sensitive to building stimuli: evidence and implications. Neuron 1998, 21:373-383. 32. Epstein R, Kanwisher N: A cortical representation of the local visual environment. Nature 1998, 392:598-601. 33. Chao LL, Martin A, Haxby JV: Are face-responsive regions selective only for faces? Neuroreport 1999, 10:2945-2950. This paper provides evidence for overlapping neural substrates in the fusiform gyrus and superior temporal sulcus for representing animals and human faces. See [34] for a contrasting view. 34. Kanwisher N, Stanley D, Harris A: The fusiform face area is selective for faces not animals. Neuroreport 1999, 10:183-187. 35. Caramazza A, Shelton JR: Domain-specific knowledge systems in the brain: the animate-inanimate distinction. J Cogn Neurosci 1998, 10:1-34. 36. Ishai A, Ungerleider LG, Martin A, Schouten JL, Haxby JV: Distributed representation of objects in the ventral visual pathway. Proc Natl Acad Sci USA 1999, 96:9379-9384. 37. Martin A, Wiggs CL, Ungerleider LG, Haxby JV: Neural correlates of category-specific knowledge. Nature 1996, 379:649-652.

49. Moore CJ, Price CJ: A functional neuroimaging study of the variables that generate category-specific object processing differences. Brain 1999, 112:943-962. This paper provides evidence for the crucial role of the left posterior middle temporal cortex in naming and matching pictures of tools. Results also indicate that processing natural objects maximally activates bilateral regions of the anterior temporal and right posterior middle temporal cortices. The results are interpreted as support for the perceptualfunctional theory of category specificity and the theory that natural objects increase demands on object processing and semantic differentiation. 50. Oram MW, Perrett DI: Responses of anterior superior temporal polysensory (STPa) neurons to biological motion stimuli. J Cogn Neurosci 1994, 6:99-116. 51. Bonda E, Petrides M, Ostry D, Evans A: Specific involvement of human parietal systems and the amygdala in the perception of biological motion. J Neurosci 1996, 16:3737-3744. 52. Puce A, Allison T, Bentin S, Gore JC, McCarthy G: Temporal cortex activation in humans viewing eye and mouth movements. J Neurosci 1998, 18:2188-2199. 53. Kourtzi Z, Kanwisher N: Activation in human MT/MST by static images with implied motion. J Cogn Neurosci 2000, 12:48-55. Together with [54], this paper suggests that the brain regions involved in the visual analysis of motion are also engaged in processing implied dynamic information from static images. 54. Senior C, Barnes J, Giampietro V, Simmons A, Bullmore ET, Brammer M, David AS: The functional neuro-anatomy of implicit motion perception of representational momentum. Curr Biol 2000, 10:16-22. 55. Hoffman EA, Haxby JV: Distinct representations of eye gaze and identity in the distributed human neural system for face perception. Nat Neurosci 2000, 3:80-84. 56. Grabowski TJ, Damasio H, Damasio AR: Premotor and prefrontal correlates of category-related lexical retrieval. Neuroimage 1998, 7:232-243. 57. Chao LL, Martin A: Representation of manipulable man-made objects in the dorsal stream. Neuroimage 2000, 12:478-484.

38. Perani D, Cappa SF, Bettinardi V, Bressi S, Gorno-Tempini M, Matarrese M, Fazio F: Different neural systems for recognition of animals and man-made tools. Neuroreport 1995, 6:1637-1641. 39. Perani D, Schnur T, Tettamanti M, Gorno-Tempini M, Cappa SF, Fazio F: Word and picture matching: a PET study of semantic category effects. Neuropsychologia 1999, 37:293-306. Data from two PET studies (with pictures and the written names of animals and tools) suggest that different brain networks subserve the identification of these categories. In particular, the results support an association between tool concepts and activation of the left middle temporal gyrus. 40. Tranel D, Damasio H, Damasio AR: A neural basis for the retrieval of conceptual knowledge. Neuropsychologia 1997, 35:1319-1327. 41. Martin A, Ungerleider LG, Haxby JV: Category-specificity and the brain: the sensory-motor model of semantic representations of objects. In Category Specificity and the Brain: The Sensory-Motor Model of Semantic Representations of Objects. Edited by Gazzaniga MS. Cambridge, MA: MIT Press; 2000:1023-1036. 42. Ishai A, Ungerleider LG, Martin A, Haxby JV: The representation of objects in the human occipital and temporal cortex. J Cogn Neurosci 2000, 12:35-51. 43. Gauthier I, Tarr MJ, Moylan J, Skudlarski P, Gore JC, Anderson AW: The fusiform face area is part of a network that processes faces at the individual level. J Cogn Neurosci 2000, 12:495-504. 44. Kreiman G, Koch C, Fried I: Category-specific visual responses of single neurons in the human medial temporal lobe. Nat Neurosci 2000, 3:946-953. This is a compelling account of category-specific visual responses from single neurons in the human medial temporal cortex. 72% of visually responsive neurons responded selectively to faces, natural scenes and houses, famous people or animals. Most of the selective neurons responded only to a single stimulus category. These same neurons responded when the patients imagined the objects [83]. 45. OCraven KM, Kanwisher N: Mental imagery of faces and places activates corresponding stimulus-specific brain regions. J Cogn Neurosci 2000, 12:1013-1023. 46. Ishai A, Ungerleider LG, Haxby JV: Distributed neural systems for the generation of visual images. Neuron 2000, 28:979-990. 47. Martin A: Organization of semantic knowledge and the origin of words in the brain. In Organization of Semantic Knowledge and the Origin of Words in the Brain. Edited by Jablonski NG, Aiello LC. San Francisco, CA: California Academy of Sciences; 1998:69-88.

58. Grafton ST, Fadiga L, Arbib MA, Rizzolatti G: Premotor cortex activation during observation and naming of familiar tools. Neuroimage 1997, 6:231-236. 59. Grafton ST, Arbib MA, Fadiga L, Rizzolatti G: Location of grasp representations in humans by position emission tomography: 2. Observation compared with imagination. Exp Brain Res 1996, 112:103-111. 60. Gerardin E, Sirigu A, Lehricy S, Poline JB, Gaymard B, Marsault C, Agid Y, Le Bihan D: Partially overlapping neural networks for real and imagined hand movements. Cereb Cortex 2000, 10:1093-1104. 61. Rizzolatti C, Camarada R, Fogassi L, Gentilucci M, Luppion G, Matelli M: Functional organization of inferior area 6 in the macaque monkey. II. Area F5 and the control of distal movements. Exp Brain Res1988, 71:491-507. 62. Hoshi E, Tanji J: Integration of target and body-part information in the premotor cortex when planning action. Nature 2000, 408:466-470. 63. Sergent J, Ohta S, MacDonald B: Neuroanatomy of face and object processing. Brain 1992, 115:15-36. 64. Leveroni CL, Seidenberg M, Mayer AR, Mead LA, Binder JR, Rao SM: Neural systems underlying the recognition of familiar and newly learned faces. J Neurosci 2000, 20:878-886. Results in this paper show that, relative to recognizing recently encoded faces, recognizing famous faces produces bilateral activations in anterior temporal and mesial temporal regions. This pattern of activation is consistent with other studies using famous faces. 65. Gorno-Tempini ML, Price CJ, Josephs O, Vandeberghe R, Cappa SF, Kapur N, Frackowiak RSJ: The neural systems sustaining face and proper name processing. Brain 1998, 121:2103-2118. 66. Damasio H, Grabowski TJ, Tranel D, Hichwa RD, Damasio AR: A neural basis for lexical retrieval. Nature 1996, 380:499-505. 67. Gauthier I, Anderson AW, Tarr MJ, Skudlarski P, Gore JC: Levels of categorization in visual recognition studied using functional magnetic resonance imaging. Curr Biol 1997, 7:645-651.

48. Haxby JV, Ishai A, Chao LL, Ungerleider LG, Martin A: Object-form topology in the ventral temporal lobe. Trends Cogn Sci 2000, 4:3-4.

Semantic memory and the brain: structure and processes Martin and Chao

201

68. Kosslyn SM, Thompson WL, Kim IJ, Alpert NM: Topographical representations of mental images in primary visual cortex. Nature 1995, 378:496-498. 69. Damasio AR: Time locked multiregional retroactivation: a systems level proposal for the neural substrates of recall and recognition. Cognition 1989, 33:25-62. 70. Gabrieli JDE, Poldrack RA, Desmond JE: The role of left prefrontal cortex in language and memory. Proc Natl Acad Sci USA 1998, 95:906-913. 71. Wagner AD: Working memory contributions to human learning and remembering. Neuron 1999, 22:19-22. 72. Poldrack RA, Wagner AD, Prull MW, Desmond JE, Glover GH, Gabrieli JDE: Functional specialization for semantic and phonological processing in the left inferior frontal cortex. Neuroimage 1999, 10:15-35. This paper provides an extensive review of the literature, as well as new functional magnetic resonance imaging (fMRI) data, demonstrating a distinction between left prefrontal regions involved in semantic processing and those involved in phonological/lexical processing. 73. Gabrieli JDE, Desmond JE, Demb JB, Wagner AD, Stone MV, Vaidya CJ, Glover CH: Functional magnetic resonance imaging of semantic memory processes in the frontal lobes. Psychol Sci 1996, 7:278-283. 74. Demb JB, Desmond JE, Wagner AD, Vaidya CJ, Glover GH, Gabrieli JDE: Semantic encoding and retrieval in the left inferior prefrontal cortex a functional MRI study of task-difficulty and process specificity. J Neurosci 1995, 15:5870-5878. 75. Wagner AD, Koutstaal W, Maril A, Schacter DL, Buckner RL: Task specific repetition priming in left inferior prefrontal cortex. Cereb Cortex 2000, 10:1176-1184. This paper provides evidence that the anterior and posterior regions of the left inferior prefrontal cortex are functionally distinct, and that priming in these regions does not derive from mere perceptual recapitulation.

76. Dapretto M, Bookheimer SY: Form and content: dissociating syntax and semantics in sentence comprehension. Neuron 1999, 24:427-432. 77. Thompson-Schill SL, DEsposito M, Aguirre GK, Farah MJ: Role of left inferior prefrontal cortex in retrieval of semantic knowledge: a re-evaluation. Proc Natl Acad Sci USA 1997, 94:14792-14797.

78. Thompson-Schill SL, Swick D, Farah MJ, DEsposito M, Kan IP, Knight RT: Verb generation in patients with focal frontal lesions: a neuropsychological test of neuroimaging findings. Proc Natl Acad Sci USA 1998, 95:15855-15860. 79. Thompson-Schill SL, DEsposito M, Kan IP: Effects of repetition and competition on activity in left prefrontal cortex during word generation. Neuron 1999, 23:513-522. This work provides evidence that activity in the left inferior prefrontal cortex is associated with increased selection rather than increased retrieval demands. These results are consistent with the hypothesis that a region of the left inferior frontal gyrus subserves the selection of semantic knowledge among competing alternatives. 80. Fiez JA, Balota DA, Raichle ME, Petersen SE: Effects of lexicality, frequency, and spelling-to-sound consistency on the functional anatomy of reading. Neuron 1999, 24:205-218. 81. Mummery CJ, Patterson K, Wise RJS, Vandenberghe R, Price CJ, Hodges JR: Disrupted temporal lobe connections in semantic dementia. Brain 1999, 122:61-73. Results from this work indicate that semantic dementia patients with extensive anterior temporal lobe atrophy still show normal or even heightened activity in subregions of the left anterior temporal lobe, but fail to activate the ventral and lateral regions of posterior left temporal cortex. These data provide converging evidence from normal and brain-damaged subjects for the importance of the anterior temporal lobe in retrieving and manipulating semantic information. 82. Petrides M: Dissociable roles of mid-dorsolateral prefrontal and anterior inferotemporal cortex in visual working memory. J Neurosci 2000, 20:7496-7503. 83. Kreiman G, Koch C, Fried I: Imagery neurons in the human brain. Nature 2000, 408:357-361.